Contemporary computers, cellular telephones, radios, televisions, and other electronic equipment are constructed using semiconductor microprocessors, integrated circuits, memory chips, and the like. These semiconductor components, which are characteristically fabricated on a semiconductor substrate, are constructed using various microelectronic devices such as transistors, capacitors, diodes, resistors, and so forth. Each microelectronic device is typically a pattern of conductor, semiconductor, and insulator regions formed on the semiconductor substrate.
The electronics industry continues to strive for increasingly higher-powered and higher-functioning circuits by device miniaturization and by creating multifunction devices on a single semiconductor chip or die.
Very large-scale integrated circuits on small areas of silicon wafers are manufactured through a series of steps carried out in a particular order. The main objectives include obtaining a device that occupies as small an area as possible and consumes low levels of power using low voltage supply levels, while performing at speeds comparable to speeds realized by much larger devices.
An important part in the circuit construction and manufacture of semiconductor devices concerns data storage, or semiconductor memories; the circuitry used to store digital information. The construction and formation of such memory circuitry typically involves forming at least one storage element and forming associated circuitry designed to access the stored information. In applications where circuit space, power consumption, circuit speed, and reliability are primary design goals, the construction and layout of memory devices can be very important.
Dynamic read/write random-access memory (“DRAM”) is a type of volatile memory in which the data stored at each location is periodically refreshed by reading it and then writing it back again to the same location, or else it disappears. Static read/write random-access memory (“SRAM”) is a type of volatile memory in which the data, once it is written to a memory location, remains stored there as long as power is applied to the memory chip (unless, of course, the data is deliberately changed by replacing it with new data).
SRAM and DRAM often compromise one or more of the primary design goals of smaller circuit space, lower power consumption, and faster circuit speed. For instance, some SRAMs include circuit structures that compromise at least one of these primary de sign goals. An example is a conventional SRAM based on a four-transistor (“4T”) cell, or a six-transistor (“6T”) cell, that has four cross-coupled transistors or two transistors and two resistors, plus two cell-access transistors. Such cells have the advantage that they are compatible with mainstream complimentary metal oxide semiconductor (“CMOS”) technology, consume relatively low levels of standby power, operate at low voltage levels, and perform at relatively high speeds. However, the 4T and 6T cells are conventionally configured using a large cell area; and this large area significantly and undesirably limits the maximum density of such SRAMs.
Other SRAM cell designs are based on negative differential resistance (“NDR”) devices. These usually consist of at least two active elements, including the NDR device. The structure and operating characteristics of the NDR device are particularly important to the overall performance of this type of SRAM cell. A variety of NDR devices has been introduced, ranging from a simple bipolar transistor to complicated quantum-effect devices. The biggest advantage of the NDR-based memory cell is the potential of having a cell area smaller than that of 4T and 6T memory cells because of the smaller number of active devices and interconnections needed in the NDR design.
Conventional NDR-based SRAM cells, however, have many problems that have inhibited their use in commercial SRAM products. Some of these problems include: high standby power consumption due to the large current needed in one or both of the stable memory states of the cell; excessively high or excessively low voltage levels needed for the cell operation; stable states that are too sensitive to manufacturing variations and provide poor noise-margins; limitations in access speed due to slow switching from one stable state to the other; and manufacturability and yield issues due to complicated fabrication processing.
One NDR device is the “thyristor” (from the Greek thyra, meaning “door”, which suggests something that is either open or closed, and thus either on or off). Thyristors are widely used in power switching applications because the current densities carried by such devices can be very high when in their “on” state.
A thyristor is a four-layer semiconductor device consisting of alternating N-type and P-type semiconductor materials (i.e., “NPNP”), with three P-N junctions. Thyristors usually have three electrodes: an anode, a cathode, and a gate (or control electrode).
A thyristor can be turned on by an initial current at the gate, and once it is turned on it then does not require any more control (gate) current to continue to conduct. Instead, it will continue to conduct until a minimum holding current is no longer maintained between the anode and cathode, or until the voltage between the anode and the cathode is reversed.
A thyristor can thus switch or control large amounts of power using but a small triggering (or control) current or voltage. Thyristors, then, act like a semiconductor analog of a mechanical switch—the two stable states are “on” and “off,” with nothing in between. Thyristors are used, among many applications, in motor speed controls, light dimmers, pressure-control systems, and liquid-level regulators.
For use in electronic applications, however, a significant difficulty with thyristor devices is that once switched to their on state, they remain in this state until the current is reduced below the device holding current. Also, in general, when the main thyristor current is interrupted (to turn the thyristor off), the time required for the thyristor to return to the off state is largely determined by the lifetimes of the current carriers in the device, and in electronic terms this can be quite long.
This inability to switch the device off without interrupting the current, and the associated slow switching speed, are significant problems in many applications and have resulted in many attempts to modify the device structures so that they can be actively and rapidly switched off.
A prior design directed to these concerns appears in U.S. Pat. Nos. 6,229,161 and 6,448,586 (both to Nemati et al.). The descriptions are directed to capacitively-coupled NDR devices such as thyristors, and to circuit applications for such structures. These devices are described as having advantages for use in designs that need NDR devices having improved on/off switching speeds and a low holding current when in the on state. They are said to be unlike many NDR devices such as conventional thyristor structures that slowly turn off due to the saturation of their junctions in the on state, and/or which may not turn off at all until the current is reduced below the holding current. Instead, the devices in these patents are directed to capacitively-coupled thyristor devices that quickly switch between a current-passing mode and a current-blocking mode in response to a capacitively-coupled activation signal that is presented adjacent a particular region of the capacitively-coupled NDR device. The switch or change from one state to the other occurs using a relatively low voltage, and the devices can be implemented in a relatively small area.
A thyristor-based random access memory (“TRAM”) array having a plurality of such TRAM cells could have a density equivalent to that of DRAM arrays and a speed equivalent to that of SRAM arrays. Hence, such a TRAM array could provide the advantages afforded by both SRAM and DRAM arrays. These advantages make such a TRAM attractive for possible use in future generations of high speed, low-voltage, and high-density memories.
However, the TRAM cell disclosed in these patents presents several major drawbacks. For example, the entire thyristor is formed in a vertical silicon pillar, requiring epitaxial processing to form the vertical configuration. This is not compatible with conventional bulk CMOS processing.
In addition, the thyristor SRAM cell (TRAM) sits in a P− well in a bulk silicon substrate. This results in a large difference in elevation between the contact on the top P+ region of the thyristor and other contacts in the device.
Also, additional masking, implantation and diffusion steps are needed to form the bottom N+ region of the thyristor.
There is also potential incompatibility with salicidation. Additionally, difficulties can arise in controlling the dimensions of the vertical pillar and reproducing these dimensions for each TRAM cell in a TRAM array.
Also, due to the existence of a vertical thyristor in each TRAM cell, the TRAM cells as a whole are not planar and are therefore difficult to scale to larger configurations.
Additionally, since each TRAM cell is fabricated prior to or after fabricating other devices (such as positive-channel metal-oxide semiconductor (“PMOS”) and negative-channel metal-oxide semiconductor (“NMOS”) supporting devices) that are connected to it, extra fabrication steps and increased manufacturing costs are incurred.
Thus, while there is a growing trend to realize an SRAM cell by using the TRAM combination of an NPNP junction connected to an NMOS transistor for ultra-high cell density, the incompatibility with conventional bulk CMOS processing makes such embedded SRAM design difficult.
Solutions to problems of this sort have been long sought, but have long eluded those skilled in the art.